The article, drawn

The lead read Who owns what in AI↗ and asked for it to show what it says (D21↗): the interconnection, the blast radius as an employee connects the assistant and as they disconnect it, the flows played out, and the evidence.

• Three figures, drawn by the page from its own data. An inline script builds them from the article's eight rows, five risks, thirteen roles and six weeks. The page loads nothing from anywhere and sends nothing.
• The blast radius can be played or changed. Play the six weeks, or connect and disconnect the mail, the CRM and the calendar, add the four boundaries, and add the instructions. A risk lights when every fact it needs holds and none is bounded, drifts to its holder, moves up after a week of silence, and ceases citing the ABP version that ended it. The counts under the figure are recomputed from the record and match the article: unbounded excess four, then three, then two. Hover a role to read the exposure in its words.
• The instruction switch shows what an instruction is. Tick it and the mandate becomes precise about the business process, every row in the gap gains an expectation, and unbounded excess does not move. The mail scope fact is cited from our Gmail record: three consent lines, each for the whole account, none for one folder, one customer or one business process.
• The six weeks as a strip: one line per risk, who held it, in which state, and the version that ended it.
• The ontology, drawn: eight node types and thirteen verbs, each readable from both ends.

What was not done:

• Blast radius as a number. The figure shows which roles a risk reaches and who holds it, and does not size the consequence, because the article scores nothing.
• Real grants. The figure runs on the article's invented agent. Wiring it to a published ABP vault is the next step.
• Evidence flowing downward. A ceased risk cites its version in text; the figure does not yet animate the evidence travelling back.